Sunteți pe pagina 1din 28

CoE9Prlig?tion (3rd ed.

'
f rom: M. I(napp & J. Dary (Eds.), ttSrylbookot Interpefso9al
pp. lO2-I29>. Newbury Parkl CA: Sage' 7Oc>>

DiscourseAnalysis

JANET BEAVIN BAVELAS


C H R I S T I N E ,K E N W O O D
BRUCE PHILLIPS

n this chapter, we first define discourse and original methods of discourse analysis
analysisand illustratethe breadthof its use- that suit their particular interestsand goals,
fulnessas a researchmethod. \Ue also empha- guided by explicit theoretical assumptions.
size how choosing to use this method can be
part of a larger changeof rheoreticalperspec-
tive. lVe devote the rest of the chapter to one
WHAT IS DISCOURSEANALYS$?
of the main purposesof a handbook such as Discourse analysis is the systematic sudy of
tlis one, which is to assistreadersin the useof naturally occurring (not hyporhetical)comrnu-
the method; thus we include many hands-on nication in the broadest sense,at the level of
issues,such as choicesabout data, recording, meaning (rather than as physical acts or
transcription, and reliability. ln reviewing features).However, a survey of the literature
theseissues,we focus on principles that might on discourse analysis would quickly reveal
be useful for helping readersto develop new that, although some r€searchersemploy the

AUTHORS' NOTE: We acknowledge the contributions of Dr. Linda Coatcs to the planning of this chapter, the Social
kiences and Humanines RescarchCouncil of Canada for rescarchgrants to the first author, and our continuing rescarch
team for sharing our daily immersion in theseissues,

102
Discourse Analysis 103

terrn to describea particular kind of analysis, Defining Discourse


it is also a label that has widespreadusage
across several disciplines with diverse goals. In new disciplines,familiar terms often take
Consequently,it is more accurateto think of on new specializedor professional meanings
discourse analysis as a cluster of methods rhat differ from their commonly usedinformal
and approacheswith some substantial com- or everyday definitions. Discourse analysis is
rnon interestsrather than as a single, unitary no different.Van Dijk ll997c) points out that
technique, the term discourseis commonly used to refer
As Vood and Kroger (2000,p. 18)explain, to a particular form of languageuse (e.g.,
rhe existence of several kinds of discourse public speeches)or more generally to spoken
analysisis undoubtedly due to the developing language or ways of speaking, such as the
nature of the field as well as to its diversedis- "the discourse of former President Ronald
ciplinary origins. Discourse analysisbegan in Reagan."Another informal usagerefersto the
branchesof philosophy, sociology,linguistics, ideasor philosophiespropagatedby particular
and literary theory, and it is continuing to peopleor groups of people (van Dijk, 7997c,
develop in additional disciplines such as pp. 1-2). In this usage,van Dijk notes, the
anthropology,communication,education,and actual languageused by a person or personsis
psychology.We find this newnessand diver- ignored in favor of a focus on the ideas or
siry a positive feature of the field. It is not philosophies expressed,According to van
hound by any single discipline, which means Dijk, the more specializedor professionaldefi-
that there is a rich infusion of ideas and nition of discourseincludes a particular focus
methodsacrossdisciplines.Nor is it commined on the actual languageused in a communica-
ro traditions of the past;indeed,many discourse tive event. A discourseanalyst is essenrially
analystsare rebelsand innovators within their interested in "wbo useslanguage, hou, why
own home disciplineswho have moved out to and uthen" (van Dijk, 1997c, p. 3). So, for
toin other like-mindedresearchers. example, a discourse analyst might examine
It is intriguing that one of the original talk occurring during encounterswith friends,
meaningsof the verb discoursewas "to travel phone calls, iob interviews,doctor's visirs,and
rcrossa courseor terrain." !0e like to think of so forth.
Jiscourseanalysisas still doing that, traveling Van Diik (1.997c)also toucheson another
rc{oss many disciplines,often into new terri- imponant distinction when he points out that
tory, rather than stayingin one place.In that languagecan be spoken, wrirten, or printed.
'pirit, in this chapterwe travelthrough (but by Each kind of languageuse, he nores, has dis-
no means claim hegemonyover) the many tinct propercies;for example,the communica-
Jomains of discourseanalysis and a wide tion may be passive(asin when an individual
variet/ of other territories where researchers readsa newspaper),more ective (as in e-mail
srudy naturally occurring language,including communication), or fully active (as in face-to-
ionversation analysis, microanalysis, ethno- face dialogue). It is important that researchers
eraphy, some areas of nonverbal communi- considerthesecharacteristicswhen conducting
carion, and mediated communication. fu is analyses.Van Diik's inclusion of both spoken
.rppropriateto this kind of intellectualinterna- and written forms of languageuse is a fairly
rionalism, the stimulus for our survey of defi- common view (e.g.,Gilben 6c Mulkay, 1984;
nrrionsand approachesis a traveler'scuriosiry Potter 6( Wetherell, 7987, p. 7; rDfood &
rbout how people do things differently rather Kroger, 2000, p. 19). It is not the case,how-
:han any goal of standardization. ever, that all discourse analysts agree. Some
t04 PERSPECTIVESON INQUIRY

would reservethe term discoursefor spoken Kinds of Discourse Data


language and use the term text to describe Another way to define discourseis to illus'
wrinen or printed language; others use the trate some of the many possible sources of
two lerms interchangeably.Most researchers data for discourseanalysis. Face-to-facedia-
define discourseas the activitiesof speakingor logue occurs in families, in most workplaces
wriring. but there are some researcherswho and public places,in psychotherapy,in courr-
include other aspects of communication in room s€ttings (e'g., examination and cross-
their definitionsas well, Brown (1995)studied examinadon), in police interviews, in medical
the role of listeners as they parricipate in examinadons and interviews, on social occa-
dialogues,and Kroger and Wood (2000)point sions,in classrooms,in meetings,and some in
out rhat some theorists, such as Fairclough psychology exPeriments'to name iust a few
(1993) and Hac€ (1995),haveeftendedthe
'semiotic seftings,l
definrtion of discourse to include lndividuals communicatein writing through
practicein other semioticmodalities"(p. 19). memos and lefters, when they post notices'
That is, rheseresearchershave a broader defi- when they publish books and articles, when
nition of discourse that includes not only professorswrite commentson students'exams
words (spokenor written) but also other kinds
or papers,and in many other settings.
of meaningfuicommunication,such as visual Mediated communication includes tele-
imagesand nonverbal movements(e.9.,ges- phone conversationsand communicadonvia
tures). Our researchgroup can be counted
answeringmachinesor voice mail, radio call'
among those who accept all of these broader in shows, and comPuter-mediatedforms such
definitions.
as e-mail, Internet chat rooms, and bulletin
Finally, there are some implicit agleements
boards. In the massmedia, mediated corlmu-
among discourse analysts that become signi- nication occurs through newspapers' comic
ficant in contrast to other approaches to
strips, TV talk shows, and political interviews.
language and communication. Discourse
We provide more examPlesthroughout this
analysts would always look closely at the
chapter,especiallyin the sectionson recording
actrlallanguageitself rather than ar secondary
and transcription, Undoubtedly, readers can
sourcessuchas reportsor descriptionsof what
immediatelythink of many other exampleswe
rvassaid,meant,or understood,That is, virtu-
have not mentioned here.
ally all discourseanalystswould agreewith
rhe conversationanalysts,such as Atkinson
Levels of Analysis
and Heritage (1984), who useonly data from
behaviors generated in their own context. Nunan (19931,whocomesfrom a more lin-
The observers' and participants' descriptions, guistic background,takes an evendifferent cut
lnrerpretations,and comments on conver- at the definition of discourse'He usestert to
sations are necessarilygathered in different refer to the wrinen or taped record of a com-
iontexts, for different purposes, and these munication event and discoutseto refer to the
rnevitably affect the descriptions, These interpretation of that event in the context in
reports might be analyzed as "accounts" which it occurs. For Nunan, the difference
Scott 6a Lyman, 1990) or as "remembering" betweentext analysisand discourseanalysisis
.rs opposed to "memory"; Banlea, 1932; that the former is the srudy of formal linguis-
Edwards 6< Middleton, L987), but they are tic devicesthat distinguisha text from random
rot substitutesfor the discoursethey describe. sentences,whereasthe lafter is also the study
Discourse Anelysis 105

of such devices but is conducted by the Languagecan be conceptualizedin basically


researcher wirh the intention of coming to two ways, es systetflor strrrcture, or as dri-
course, practice (praxis) or communication.
understand the purpose and function of the
If one gives priority to the former, we can
discourse as well as the context in which it
talk about a formalist(ic) framcwork; here,
developed (p. 20). What he calls discourse linguistic expressions can be treated in
analysisinvolves languageas it is being used, abctracto. Priorities to the latter yield a more
whereas text analysis is concerned with functionalist(ic) paradigm; its focus on com-
pafterns and regularities thar occur in written municative meaningsand functions makes it
language, such as phonemic or grammatical necessaryto take contextsinto account. (p. 3)
analyses. That is, discourse analysis is
concerned with panerns and regularities in Ve have already encountered a functionalist
language but also with the people using view in van Diik's (1997c\ stated interest in
language(what they mean and the purpose to "zlro uses languagc, how, uhy and when"
which language is put) and the context in (p. 3). Vith respect to our own research
which it is used. Nunan (1993) says of dis- group's approach to discourse analysis, we
course analysts, "Their uldmate aim is to take a strongly functional approach, focusing
show how the linguistic elements [found in mainly on how dialogue works and what a
languageJenablelanguageusersto communi- panicular phenomenonis doing (or how it
catein context' (p. 20). Thus Nunan's interest works) in its immediate communicative
in linguistic elements leads him to study, context.
among other things, linguistic devicessuch as So far, we have looked at discourseana-
pronoun usage and conjunctions that enable lysts who are concernedprimarily with lan-
people to build explicit relarionshipsbeween guage use, that is, at researcherswhose main
enritiesand eventsin their discoursewith each interesris in what is said or written. However,
other in different contexts(p. 57).He is also some analysts focus on the kind of discourse
interested in analyzing how smaller compo- that involves ideas or philosophies propa-
nents of languagecontribute to broader social gatedby particular people or groups of people
meanings.Thus his researchspanstwo levels (van Dijk, 7997c, pp. 1-2). Analystssuch as
in that he breaks discourse down into its Caldas-Coulthardand Coulthard ( 1995b)call
component parts and also looks at how the themselvescritical linguists and, along with
parts contribute to the formation of meaning researcherssuch as Fairclough ij992), Fowler
in socialcontexts. (L9961,and Hodge and Kress (1993), call
Although they often use different terms, their research critical discourse analysis
other researchersoften make the samekind of (CDA), which they define as "an analysisof
distinction (between analyzing parts of lan- public discourse,an analysisdesignedto get
quageversusbroader issuesof meaning)when at the ideology coded implicidy behind the
outlining their positions.Stubbs(1983)distin- overt propositions,to examine it particularly
guishesbenveenlanguageanalysishlow the in the context of social formations' (Fowler,
It'vel of the sentenceand "language above the 1 9 9 6 ,p . 3 1 .
jentenceor abovethe clause"(p. 1; quoted in In CDA, the focus is not entirely on the
khiffrin, 1994, p.23). Schiffrin (19941 pro- actual words written or spoken but also on the
posesthat the distinction is berweenformalist representationsimplicit in the words. This
or structurulist) and functiowlist views of kind of analysis,one might observe,can be far
).rnguage. Linell (1998)makesa similarpoint: abovethe level of the sentence,becauseir may
106 PERSPECTIVES
ON INQUIRY

be less concerned with what is spoken or I want to open up the field of mcanings to
wrirten and more concernedwith the broader which dirourse analysis could bc applied
message,philosophy, ideology, or idea con- beyond spokeninteraction and wrinen forms
veyed. CDA might, for example,reveal thar by saying that we find discoursesat work in
particular views of genderor racemisrepresent terts. Tex6 are delimited tissuesof meaning
reproducedin any form that can be given an
or disron characterisricsof the people repre-
interpretativegloss,(p. 5)
sented.The goal of critical discourseanalysrs
is to expose the misrepresenrationor distor-
He provides an example rhat is useful in clari-
tion in order to "defamiliarize"the public at
fying his meaning. He describesan electronic
largewith the negativerepresenration:
game that displaysa small moving male figure
waving a crucifix at ghosrs descendingfrom
Discourse is a maiorinstrumentof powcrand
controland CriticalDiscourse the top of the screen to their graves. Each
Analysts. . .
fecl that ir is indeedpan of their profcs- ghost that is prevented from landing by the
sionalrole to investigate, revealand clarify crucifix-waving man is consumed in flames,
how power and discriminatoryvalue are and the player is awarded 10 points. This, in
inscribedin and mediatedthroughthe lin- Parker'sview, is a text that conveysChristian
guisticsystem: CriticalDiscourseAnalysisis discourse. (Interestingls we can also see by
essenriallypoliticalin intentwith its practi- Parker'sexamplethat he is another researcher
tioners acting upon rhe world in order
who extends the meaning of discourse to
to transformit and therebycreatea world
wherepeopleare not discriminated include means of communication in addition
against
bccause of sex,colour,creed,ageor social to written and spoken languageuse.)
class. (Caldas-Coulthard 6c Coulthard.
7995a,p. xil
Language and Reality
Parker (79921,a socialpsychologisr,advo- One funher imporrant difference musr be
cates a similar kind of critical analysis. His menrioned.Discourseanalystsclearly differ on
particular interestis on the role of discoursein how they conceiv€ rhe relation berween lan-
the reproduction and transformation of mean- guageand realiry.'We usethe terrnsrealist and
ing. Discourses "both facilitate and limit. antirealist ro refer to rwo contrastingpositions
enable and constrain what can be said (by with respectto their subject maner. In stating
whom, where,when)" (p. xiii). Parkerdefines the distinction so briefly, we identify the cru-
discourseas 'sets of statementswhich consti- cial issue;however, as is often the case with
rute an objec" (p. 3; seealso p. 5). His goal, brevity,we risk oversimplifyingat the cost of
like that of Caldas-Couhhardand Coulthard, the loss of finer distinctions. Here we aftempr
is unequivocally emancipatory.That is, he only to introduce some of the issuesinvolved.
endeavorsto revealproblematicviewsas such. Some analysts, such as critical discourse
He statesrhe casequite strongly, mainraining analysts among many others, are realists in
that an "amoral/apoliticalpsychologyis worse that they maintain that cenain descripdonsor
than useless"(p, 2). constructionsof realiry are more accuratethan
I(rithin rhe writings of crirical discourse and preferableto others, even if nor perfectly
analysts, the rerm text has a different (and accurate and universally valid. lmponant in
more complex) meaning frorn the ones dis- this view is the idea that it is possible to
cussed above, which usually define it as replace unfair, inaccurate constructions or
written recordsof discourse.As Parker(1,9921 representationswith fairer and more a@urate
explains: represenrations.Note that only if there is a
Discourse Analysis t07

real realiry can any representationof it be seen identify the meaningcorrectly. But movements
f,s accurat€or inaccurate, have no inhcrent, essential meaning; rather,
The poinr at which this topic becomes they can be given multiple meanings by
different interpreters (and by the same inter-
;omplex and somerirnes very confusing is
preter on different occasions),meaningsthat
when we a$empt to answer the question,
can vary acrosssituations, (Wood & Kroger,
\t(4rat is real? Most theorists (but not all) 2Q00,p. 12)
rvould agree that the physical world exists.
However, theorists vary on whether the social
practicesand institutions createdby humans- Having briefly described the range of realisr
rvhich are therefore social constructions and and antirealist positions, we would like to

rvould not exist without human participa- clarify our own position here. First, it is
rion--should be consideredreal in the objec- important to distinguish berween physical
rive sense.Funher, theoristsalso differ on how eventsor objects in the world and the mean-
languageis related to reality, whether physical ingsof thoseeventsor objects.Supposesome-
or social. one opens a door in the presenceof another
Analysts such as Poner (1995) and Willig person.It is possibleto say that this event is
rl999l are antirealists, who maintain that real in the sensethat a movementoccurredand
reality in the commonly undentood meaning the door is now open. However, this move-
of the term does not exist. IVhat exisr are ment can take on almost any meaning,
descriptions, constructions, and repres€nta- depending on context and interpretation: It
rions that cannot be judged to be either true or could mean that the person who opens the
false.In explainingher position,\fil[g (1999) door is declaring an intention to carry out a
proposes: requestederrand or an intention ro leave the
relationship; that he or she is just letting in
[These]discourseanalystsconceptualize fresh air or implicitly complaining about
languageas consdtutiveof experience rather smoking; that he or she is advenising that the
than representationalor reflective[of experi- ensuing conversation is to be public rather
ence].They argue that the linguisticcate- than "behind closeddoors"; and so forth. In a
gorieswe usein order to "describe"realiry
great deal of social life, such interpretations
are not in fact reflectionsof intrinsic and
are elastic,as the antirealistspropose.The rwo
defining featuresof entities.Instead,they
bring into being the obiects they describe. individuals in the room can disagree about
Furthermore,thereis alwaysmorethan one what the act means,and either of them can
way of describingsomethingand our choice change(or lie about) what he or she says it
of how to usewordsto packageperceptions means.However, we would point out rwo sig-
and experiences givesrise to particularver- nificant qualifications on this elasdcity. First,
sionsof eventsandof reality,It is in thissensc there are occasionswhen cenain meaningsof
that languagecanbesaidto constructreality,
eventsare nor negoriable.lf the door is a hatch
(p. 2; emphasisadded)
on a deeply submergedsubmarine,then open-
In this view, realiry is itself a construction ing it means,among many other things, that
made in language.Crucially lacking, antireal- unprotected occupants are (really) going to
ists claim, are any universal standards by die. Second,for a great deal of what we do in
rvhich to judge the veracity of any particular everydaylife, there is substantialsocial con-
representationor conscnrction.For example: sensusabout meaning.Opening a door ahead
of another person who is walking toward it
The (conventionall
assumption
is tharmean- usually meansholding the door open for that
ing residesin movemenm-wejust needto person, and the other is very likely to go
r08 PERSPECTIVESON TNQUIRY

through the doorway first. The possibility that describesor constructs different versions of
the act could sometimesbe a trap or a ioke something real that exists independent of
does not obviate the higher likelihood that it language.The meaningsgiven can be more (or
was an act of courtesy; indeed,that very con- less)accurate.Others maintain that there are
sensusmakesthe trap or ioke possible.Ve are only versionsof reality; it is languageitself that
able to navigate social life becauseshereis a createsand constitutesrealiw.
great deal of consensusabout meaning;if this
were not so, almostno socialaction would be
NEW PERSPECTTVES
possible.Perhapsone of our tasksshould be to
explore such consensusas a ropic in itself. Discourseanalysiscan be a method that com-
We can extend the same reasoning to a plements existing communication research
discursive example: Iflhen a witness takes an tools (questionnaires, interviews,ratings,and
oath in a courtroom, this discursive act so on), to be used within traditional dreoreti-
changesthe statusof subsequenttestimooy.If cal ftameworks. Or it can be pan of a zub-
the witnesslies, he or she can then be prose- stantially different approach with several new
cuted and imprisoned for perjury, a penalry theoretrcal and methodological premises. Ve
that does not exist in other senings.Therefore descrih these alternative premises in rhis
the oath, its meaning, and its consequences section,noting thag although our researchgroup
are, we could argue,effecrivelyreal although has eventually adopted all of them, readers
clearlya socialconsrrucrion. may wish to considersomebut not others.

Summary The Primacy of Discourse


The reader can seeby now that rhe terrain A significant theoretical choice can be to
of discourseanalysisis varied but has some focus on the discourse itself rather than on
common feaores. First, there are differences inferred intrapsychic (mental) processes,such
berween what different analysts rreat as dis- as cognition, emodon, modvation, abilities,
course, These range from just written (or just goals, intentions, attitudes, and personal-
spoken)languageto other modesof communi- icy characteristics. Communication is often
cation, such as nonverbal gestures.As well, treated as secondary to rnental processes,in
discoursecan be used to describea particular one of two ways. First and most obviously,
ideology or philosophy that is implicit in dif- communicativeacts may be of interest simply
ferent forms of languageuse. becausethey provide a means of revealing or
It is not surprisingthat approachesro analy- studying mental processes.For example,
sis also differ. Some researcherslook at lan- answers on questionnairesand resPonseson
guagefrom a more strucnrral or formal point radng scalesare typically seenas direct indica-
of view in order to analyze,for instance,the rions of anirudes and preferences,not as dis-
linguistic devicesthat can be seento constiilte course. Similarly, interviews are usually seen
language. Other researchersrake a rnore as a way of finding out what is on inter-
functional approach, looking at languageuse viewees'mindsrather than as dyadic commu-
in its social coDtext-for example, examining nication ber,weeninterviewer and interviewee.
questions about who uses language,when, And many if not most studiesof facial expres-
how, and for what puposes. sion derive from a primary interest in emotion
Finally, an imponant issuein the field is the (e.g.,Ekman, 1993; but see also Bavelas&
relation benryeenlanguage and reality. Some Chovil, t997; Kraut & Johnston, 1.979).
discourse analysts mainrain chat language Emotions are not directly observable, so the
Discourse Analysis 109

srudy of facial expression is a preferred A. That tree has, uh, uh


behavioral method for inferring them. Used in
B. Tentworms.
theseways, discourseis a behavioralmeansto
intrapsychicends.The alternativeis to become A. Yeah.
anracted to the discourseitself, treating it as B . Y e a h .( p . 6 )
intrinsically interesting and wonh analyzing
in detail. Sacks (quoted in Jefferson, 7973, p.59) pro-
The second way in which researcherscan vides another example:
treat discourse as secondary is to explain it
by mental models. That is, even when the A. I heard you were at the beachyesterday.
discourse is the initial focus of interest, the What's her name,oh you know, the tall
redheadthar lives acrossthe streetfrom
focus quickly shifts to mencal process€srhar
Larry? The one who drove him to work
might explain it. The questionof "why" indi-
thc day his car was
viduals or groups communicate in particular
ways is most often answeredby hypothesized B. Oh Gina!
mental processes,such as their cognitions or A. Yeah Gina. Shesaid she saw you at the
personality. If the cause is intrapsychic, the beachyesterday.
discourse becomes merely an effect or by-
product and may not even be examined Clark and Wilkes-Gibbs (1986) point out
closely. (For a fuller discussionof the prob- that such excerprs are highly social and col-
lems with and alternatives ro this approach, laborative. Clark and his collaborators have
seeBavelas,1991.) conducted experiments showing how, even
We can illustrate these issues with the when one person is rhe official speaker,both
example of language production, which is individuals contribute possible terms, modify
usually seen as the outcome of the speaker's them, and work out mutually meaningful
cognitive processes.There are many elaborate references(which are not necessarilyintell!
cognitive models that explain how an indivi- gible to oursiders)(Clark 6c Wilkes-Gibbs,
dual finds a word or produces sentences,and 1986; Schober 6< Clark, 1989). Our main
of course sophisticatedmental processesmust point here is that their researchfocus is on the
be involved in this most skilled of human dialogue that produces references,not on the
activities. However, Bavelas and Coates mental processesof the speaker.
(19921 have questioned whether existing Thus one can chooseto study the discourse
mental models of language production and itself without using or invoking mental
comprehensioncan account for the observed process€sat all. This proposal is not the same
precision and complexiry of dialogue. At the as a radical behaviorism thar deniesthc exis-
very least, a thorough appreciation of the tence of mental phenomena-which clearly
nature of the discourse irself (versus hypo- exist and are importanc-bur rarher a shift of
therical examples) must be the criterion by interest to the other, equally interesting phe-
which such modelsare generatedor iudged. nomena to be found in discourse.There is no
An even greater shift is to examine the a priori reason why inferred mental pheno-
discourse of language production without mena should be given primacy over discourse.
reference to mental models. Clark and One might fairly ask, however, if re-
!flilkes-Gibbs (1986) proposed that the searchersturn away from mental models,
speaker and addresseework toterher to are they not also abandoning rheory and
choosea word or term to describesomething. becomingmerely descriptive?If individual or
For example: intrapsychicexplanationswere rhe only way
110 PERSPECTIVESON INQURY

theory could be built, then eschewingthem to find previous theory and research,(b) forming
focus on discoursewould indeed be atheoreti- a hypothesis, and then (c) going to the data
cal. However, it is also possibleto answer the (e.g., discourse) to test the hypothesis. An
question of "why" by examining hout and alternative approach that is consistent with
what for (Varzlawick, Beavin Bavelas, 6< discouneanalysisis to go "backward," that is,
Jackson, 1967, p, 45), that is, by looking at to start inducrively by studying the discourse
interpersonalprocessesand at the socialeffects first, generating hypotheses or explanations
of communicative asts. As theorists, w€ can from what one observes,and only then going
look ourward rather than inward. For example, to the literature.2 (For further discussionand
in our studiesof equivocalcommunication, we examples,seeBavelas,1987, 1995.)
focused on goals, not as mental intentions An inducriveapproachto discourseanalysis
or plans,but as propertiesof the socialworld: involvesat leasttwo new premises.The first is
the options available and their consequences a different assumption about the role of the
(Bavelas, 1991; Bavelas, Black, Chovil, & literature in research.Traditionally, research
Mullen, 1990). Specifically,we have shown extends an existing literature, building on it
that equivocationoccurswhen all of the direct study by study to ensurethe accumulation of
options (e.g.,lying or telling a hurtful truth) knowledge. However, entirely new insights
are negativeand to be avoided.In this view, and approach€scannot come from the litera-
equivocation is a way of rraversinga social ture. If the idea is already in the literature, it is
minefield rather than the product of inferred by definition not new. Moreover, the lenses
intrapsychic processes. imposed by the literature will inevitably
So far, we have suggesteda primacy of dis- constrainthe way in which one can seethe dis-
course as both the focus of interest and rhe course. Often, rheseconstraints are unrecog-
sourceof explanation. It is possiblero go a step nized becausethey are implicit in a particular
funher and adopt the even stronger prernise paradigm and therefore difficulr even to be
that discoursecreatesour socialworld. Unseen aware of, much lessto question. Without dis-
mental processesare not social. Only observ- paraging the obvious importance of previous
able actions or their consequencescan have research,we proposea balanceberweenconti-
social effects,and languageuse is arguably the nuity and innovation that values working
predominant social action out of which our both within and outside of existing topics
social lives are created.Thus it is possibleto see and frameworks.As Yngve (1970) observes,
interp€rsonal communication not merely as "There is nothing like viewing video rapes of
reflectingindividuals'viewsof the socialworld, actual communicative activity to dispel any
but as creatingand sustairungthe socialworld, preconceptions one may have" (p. 573).
as the fabric of social life itself. A discursive The most likely source of new ideas in
theoretical approach focuseson how the par- many naRlral sciencesis direct observation,
ticipants do social actions. the results of which often violate our precon-
ceivednotions:
An Inductive Approach:
Learningfrom rheDiscourse A good way to tell how the work is goingis
to listen in the corridors.[f you hear the
Another new perspectiveone can adopt is
word, "Impossible!"spokenas an expletiv,e,
an inductive stance. Most researchers,espe- followed by laughter,you will know that
cially students,have learned that one stans a someone's orderlyresearchprojectis coming
researchproiect by (a) going to the [terature m alongnicely.(Thomas,1974,p. 140)
Discourse Analysis 111

For example,Coates(1991) took an inductive unassailably true; the conclusion is not. All
approach to irony in dialogue and discovered veteranresearchershave enough experienceof
that it is a collaborative, not an individual, failed proiects or of cherishedbut consistently
phenomenon. Microanalysis of actual dis- unsupported hypothesesto know that think-
course has been a fruitful source of enrirely ing and wishing do not make it so. A great
new ideas, from the original Naural History deal of the methodological logic and machin-
of an trnterview proiect in the 1950s (keds- ery of scientific inquiry has been devised to
Hurwitz, 1987) to conversation analysis. guard against this unwanted possibiliry. Put
lndeed, given the relative paucity of data more positively, the researcherwill certainly
about the nature and workings of actual dis- choose to observe data that fit his or her
course, there is a great deal of observational interests(e.g.,dyadic or group interaction, the
work to be done. Researchersin our field may media)and will also approachthesedata with
be like the 19th-century naturalists (of whom a set of assumptions,preconceptions,and even
Darwin was only one) who faced an exciting hunches.Curiosity, which is the driving force
new world of animals and plants to carefully of research, is rarely neutral, but this need
describeand understand. not shut out new or unexpectedpossibilities.
If we begin to question the raditional role lf one is truly curious, then theoretical pre-
of the literature, we can also question the role ferences will not dictate one's observation.
of theory and hypothesis testing, which are Virrually all inductive researchersrequire a
usually seenas necessaryto guide research.ln number of instancesof a phenomenonbefore
the traditional view, one moves from general they conclude that it is more than one of
to specific; theory and hypothesiscome first, a kind.
then observation: theory and deduction -+ The second concern about conclusions
hypothesis --> confirmatory obseruation. ln derivedfrom inductiveresearchis the dangerof
contrast, observation can lead to theory by a particular kind of circularity. If the researcher
moving from the specific to the general. An is creativeenough,so the argument goes,he or
initial inductive phase may lead to exciting she will always be able to find some panem
new hypothesesand eventheories:obsentdtion that fits this particular set of data, but that
and induction --shypotbesisand theory. Some pattern may fit only thesedata becauseit was
inductive researchers,including our group, created from and for them. Again, new data
add a third stepby returning to the data to test are the answer.That is, although it is true that
the new hypothesis,in a cycle of induction and humans are very good at finding patterns and
deduction: obseruation and induction -+ can evenfind them in random numbers (or the
hypothesisand theory -+ confirmatory obser- stock market!), new observationswill not
uation, !7e return to these rwo alternative confirm a parern that fits only the limited set
approacheslater in this chapter. of data that generatedit.
There are two cofilmon and reasonablecrit- For both of these concerns, there are also
icisms of an inductive approach. The first is proceduresthat provide technical safeguards,
rhat one cannot approach data free of precon- namely, interanalyst reliability and cross-
ceptions;no researcher'smind is a blank slate, validation, to be described below. Some
to be written on by data without any prior researcherscharacterize these techniques as
ideas influencing what he or she sees. positivist and quantitative, and will not adopt
Therefore. some conclude, researcherswill them. In the next section we try to show how
seeonly what they are prepared or guided by they can be a narural extension of a logic
rheir theories to see. The first statement is shared by all researchers,of whatever stripe.
1.r2 PERSPECTIVESON INQUIRY

GETTING DOWN TO DETAILS: hypothesis-testingdesigns; here we describe


TECHNICAL ISSUES some of the more specific aspects of that
choice. Anyone exposed to typical methods
All researchis about details; the big picnrre of
courses is likely to be more familiar with
theory exists only insofar as the details of data
hypothesis-testing( "theory-driven" ) research,
createand support it. Discourseanalysisdif-
which is often described as the pinnacle of
fers somewhat in the kinds of technical issues
researchdesign.Becauseof the starusafforded
thar are important; these are quite different
it in some quarters,hypothesistesting is often
from, for example, the details of experimenul
the designthat researchers choosefirst, for any
researchusing questionnaires.Technical issues
topic. We do not subscribeto this "evolution-
such as recording and transcription replace
ary scale" view of researchdesign, especially
more familiar ones, such as the psychometric
when it relieson the natural sciencesas a role
aspects of rating scales or the intricacies of
model. A careful readingof the history of (and
factorial designs.ln this section we consider
contemporary research in) life sciencessuch
choices of researchdesign, reliabiliry, record-
as biology reveals a strong foundation of
ing, transcription,and someethicalissuc. We
painstaking inductive observation. However,
continueto emphasizethe broadertheoretical
design prejudice runs in both directions. Just
assumptionsthat underiie particular method-
as some experimentalists mistrust inductive
ologicalchoices.
research, some inductive researchers reject
experimentalor hypothesis-testingdesignsout
Research Design of hand, primarily becauseof historical posi-
tivist associarions.Our eclecticview is going to
A researchdesignis simply an orderly plan
offend extremistson both sides,so we address
for gathering and analyzing data, one that
our remarks to those who want to make up
serves the pafticular goals of the researcher
their own minds.
for a particular study. It is not necessarilyor
If we put preconceptionsand receivedwis-
even usually an experimentaldesign.What is
dom aside, what are the differencesbetrveen
crucial is that researchersgive considerable
hypothesis-testing research and inductive
forethought to the plans and goals embodied
research?In hypothesis-testingresearch, the
in their designdecisionsbeforeacringon them,
theory drivesthe designfor gatheringthe data.
so that the time and effort to be invested in
Ideally, the researcherhas a cleady formulated
carrying them out are likely to serve their
hypothesis,that is, a seriousbet about what is
purpose.In our view, researchers shouldmake
going on and why. Multivariate statisticaldata
decisionsthat fit their own goalsand interests,
dredging is nor the sameas hypothesistesting.
not any particular ideology; questioning
The researcherhas to develop a desigrr that
methodological dogma and divisions can lead
clearly anticipates all possible outcomes,
to creative new options (seeBavelas,1995).
including one that would suppon his or her
An examination of our group's research,for
hypothesisand othersthat would not. Only if
example,revealsa considerableheterogeneity
the outcome can either advance or diminish
of choices.
the plausibiliry of the researcher'stheory is
there true hypothesis testing. A hypothesis-
lnductiue Versus
testing design demands an up-front commit-
Hy p ot h esi s-Testing D esigns
ment, and a great deal of experimental,
As describedabove,one major choiceto be quantitative, statisticalresearchdo€s not meet
made at the outset is between inductive and this standard. On the other hand, contrary to
Discourse Analysis 113

a common antirealist criticism, experiments setof data inductively, the researcherlimits the
never attempt to (and cannot) "prove" that a anaiysisto a subaampleof the whole (ideally,
hypochesisis "rrue." Rather, experimentsare a randomly chosensubsample).Basedon this
one way of finding out which explanation of initial inductive analysis,the researchermakes
the data is more probable or which hypothesis a commitment to what has been found and
is the best fit-at leastundl a bener one comes then is prepared to test it in the remainder of
along. the sample. If the original pattern was ran-
In inductive research.the data lead to the dom, ir will disappear in the new data set.
theory. Inductive researchservesthe goal of a Dependingon the outcome, a cross-validation
researcherwho, at a particular point, has a strategy provides either a useful safeguard
strong inruition that somethingnew and inter- against the potential weakness of inductive
estingmay be going on, which he or sheneeds researchor a powerful demonstration of its
to learn about. This approachmay be an early validiry.
stagein the research,or it may be the continu- Returning to the choice between induqtive
ing preferenceof the researcherat any stage. and hypothesis-tesringdesigns,one possibiliry
The inductive researcher'sprimary moriva- is to use both, at different times. In our own
tions is curiosity abour the data. As noted ear- research,we often follow a cycle of, first, going
Iier, no inductive researcheris free of a general to the data to learn, then cross-validatingon a
framework of assumptions,but the ideal is to larger data set, and then experimentallytesting
Iearn from the data and be prepared to have the hypotheses developed in the inductive
one's preconceptions changed by the data. phase.The challengefor all researchersis to
New ideasor entirely new frameworks are not practice tolerance toward others who make
going to come from the library; they will come different choices.In the end, all of us must be
from carefully observeddata. There are some able to provide preciseanswersto thesegues-
guideluresfor embarking on this open-ended tions: What are you claiming? How do you
process(e.g.,Bavelas,1.987; Wood & Kroger, know?
2000), which requiresa good deal of fairh in
oneselfand one's abiliry to observecloselyand
Agreement Among Analy sts (Reliability)
to be fair.
The major objection to inductive researchis It is hard to believethat the ordinarily bor-
a concern that the researcherwill invent a ing topic of interanalyst(alsocalledinterrater,
paftern in the data whether one is there or not. interjudge, and interobserver) reliability is as
That is, the valuable creativepossibilitiesof controversial as it is in the field of discourse
this method must be weighed againstthe more analysis.Some researchersreject all discourse
mundanepossibiliry of rhe researcher'ssimply analysison thegroundsthat it is not reliableor
fooling him- or herself. However, logic "objective," whereas others reiect reliability
provides an answer that many inductive procedures as futile efforts to establish
researchersfollow, formally or informally. "ffuth." We discuss both of these posidons
Having observedand made inductive proposi- here.
dons based on observeddata, most inductive Experimental r€searchers,especially those
researchersgo on to seekconfirmadon of their in psychology,place a high value on objectiv-
conclusionsin new setsof data. One designfor iry, in the sensethat the researchermust not
formalizing this process is analogous to the influencethe measurementof events.(Readers
statistical procedure of cross-ualidation in will recognize that this is a strong realist
multiple regression.When approachinga new stance.as describedearlier.) To the extent that
114 PERSPECTIVESON INQUIRY

discourseanalysisfocuseson meaningand not epistemology--the belief that there is a reality


on physical actions, it must rely on interpreta- independent of the observer. This is not
tion by the analyst.In the view of somecritics, an uueasonable reaction to the strong pro-
all interpretation is intrinsically biased and obiectivity stance outlined above; the rwo
subjective. Becausethe researcheris making positions are mirror imagesof eachother. And
interpretations, he or she is influencing the if, as argued above, all measurementinvolves
measurement, which therefore cannot yield at least some interpretation, then the pursuit
objectivedata. of an obiective truth free of interpretation is
The answer to rhis criricism is rhat all indeed a misleading,even useless,exercise.
sensory data are interpretive, including read- We would like to articulate an alternative
ings on a dial or thermometer and counts of reason for and use of interanalyst agreement,
physical movements. Cenain measures are one that does not claim that it establishes
conventionally called obi eaiue simply because the truth of a rneasureor an interpretanon.
it is obvious that independent observersare One simple way of debunking that criticism is
very likely to agree:If each of us looks at the to point out that there is no standard of truth
same thermometer at the same time, presum- in the calculation of interanalyst reliabiliry,
ably we would have 100% agreementabout The usual calculation for qualitative data is
the temperature. However, there could be simply the number of agreements between
disagreementsdue to different viewing angles, analystsdivided by rhe number of their agree-
differences in eyesight, or just plain error. ments plus disagreements.For the rarer qwln-
Early astronomers learned that their own titadve measures,agreementis calculated by
reaction times affected rhe measuremenBthey correlation, or r, berweenanalysts' decisions,
made and had to be compensatedfor; this did but the same point holds: The analysts are
not stop astronomy from becoming accepted being compared only to €ach other, not to
as an objectivescience.Ifone acceptsthe inter- 'correct'
some independent standard. The
pretive role of the researcherin all measure- quesdon that a reliability figure answers is
ment, then the dichotomy berween obiective simply, How well do analystsagree?There is
and nonobiective measurement becomes, no claim to the truth or correcmessof what
instead, a continuum of demonstrated agree- they agree on (which is conventionally called
ment between independentobsenters,ranging ualidityl.
from 100% ro much lower or even nonexis- Stripped of the claim of truth, there is quite
tent agreement. When independent analysts a different benefit to dris kind of reliability.
agreehighly, then the fear of interpretive bias Establishing agreement among independent
is unfounded,and any such measuremust be analysts requires the careful explication of
accepted on equal footing with other, more what is being examined and how it is being
conventionally obiective measures. interpreted,That is, achievinghigh agreement
However, a different group of researchers on complex interpretations of discourse
rejects the pursuit of interanalysr agreement requirescareful and explicit description of the
for other reasons. Some simply feel that the interpretive and reasoning processes. This
original analyst must be correct, by definition, requirement is as valuable as the goal of
or that the skilled intuition of an experienced demonstratingagreementitself. In our experi-
analyst is not replicable. Others, especially ence, we always understandtlre phenomenon
antirealists,seeany aftempr to establishinter- more clearly and deeplyafter we havedone the
analyst reliabiliry as an effort to establish hard and iterative work of describing (and
the "truth,' and they reject the entire realist debating!) it sufficiently to achieveagreement.
Discourse Analysis 115

The extendeddefinitions that are necessaryfor the second step, she analyzed each of these
high agreement require clear statements of located instancesfor the kind of anribution
what the phenomenon is and is not; many, being made; a separatercliability was required
many examples; and a guided decision for this analysis.(A detailed summary of the
process.(It is often helpful to put the laner in process our group uses for developing and
the form of a decision tree, which articulates assessingthe reliabiliry of new methods of
rhe interprerive reasoningas a seriesof nested analysisis availablein Coates,1998.)
decisions.) When new analysts can follow
the processand agree,we know we are clear
about the phenomenon we are srudying, and Choosing and Recording the Data
'We
this increased clariry also helps us in our have two purposes in rhis subsection.
responsibiliry to describe it fully in research First, we want to provide readerswith several
articles. ideas about where to locate discursive data
There is a final, technical point we want that will suit their interestr-not an exhaustive
to make about interanalyst agreernent, for list, but one intended to fuel readers'imagina-
readers who decide to pursue it. ln discourse tions. Second, we discuss rhe options for
analysis,unlike other occasionsfor interrater recording the data for later analysis, which
reliability, there are often two levels of deci- necessarilydiffer depending on whether the
sion: (a) locating instanceswhere the phenom- original discourse occurred in writing, in
enon is occurring and then (b) a more specific spoken form, or face-to-face.
decision about uthat kind of instance this
one is. The first level is not necessaryin, for
Recording and Transcription
ersmple, questionnaire research,becausethe
as Tbeoretical Decisions
answer will be located where the question
was asked. However, in naturally occurring In a classic anicle, Ochs (1979) propos€s
language, the individuals structure their own that transcriptionitself is rheory; we would add
discourse, usually for reasons that are very that recording embodiestheory as well. That is,
different from the researcher'sinterest, and knowingly or unknowingly, researchersmake
much of what they say and do is not relevant imponant theoretical decisions when they
to what the researcher is focusing on at the decide how to record and transcribe their dara.
moment, So the res€archer needs, fust, to We review some of these theoretical decisions
locate sites where relevanr discourseis occur- below, as well as some practical considerations.
ring. A concrere example will illustrate this Choicesregarding how and what to record
process.Coates (1997) was interestedin the and transcribe are inevitably guided by
kinds of attributions that trial judges make assumptions and presuppositions about the
about what caqses offenders to commit phenomenonor processbeing studied. These
crirnes-+hat is, in how judges discursively assumptionsmay not consdrute a fully ani-
construct responsibiliry. The legal texts that culated theory; they may not even be recog-
Coates analyzed were the full judgrnents nized as assumptions. However, the act of
deliveredat the endsof uials, so they contained choosingto record or transcribein a panicular
a great deal of information unrelatedto anri- way inevitably makes some aspects of the
butions. Therefore her first step was to locate, discoursemore salient while ocher aspectsare
reliably, places in rhe texts where the iudges unrecordedor left on rhe curdng room floor,
were making anributions about the causes so to speak. It is therefore essentialthat the
of the crimes (e.g.,"due to his alcoholism"). In researcherbe explicit about rhese decisions,
1t5 ON INQUIRY
PERSPECTTVES

making deliberatechoicesabout what is to be and transcription can affect the data and their
included and excluded and ensuringthat these interpretation is essentialto the researcher's
choicesare guided by and consistentwith the reachingeither goal.
goals of the research.
It will becomeobvious in the following that
Textudl Material
our research group's theoretical position is
always that we should record in the same lf the researcherhas chosen to work with
mode in which the participants are interacdng, written or visual material (e.g., newspaper
on the assumprionthat they will useall of the articles, e-mail messages or chat room
featuresavailableto them in any given mode. exchanges, memos or letters, wrinen legal
That is, if one is prepared to analyze only iudgmenm, media ads depicdng nonverbal
words, then one should study messagesthat communication),it is fairly easy for him or
were originally wrinen (Bavelas, t984; her to obtain a printed copy of rhe entire text.
Bavelaset al., 1990, chap. 5). This revealsour Another potential advantageto this choice is
assumption that participants select or omit that wrinen data can often be renderedanony-
certain behaviors becausetheir receiverswill mous, so ethical issuessuch as the protection
or will not seeor hear them; for example,they of the participants'identitiesare simpler,and
will use voice quality when they know it will participantsmay be much more likely to give
be heard. but will use other alternatives in their permission for use of the data. Also,
writing. Many other researchers have differ- some written data are in the public domain
ent views of communicatorsand communica- and requireno permissionat all for their use.
tion and wouid therefore disagreewith our In approachinga printed record of wrinen
position and make other decisions; their dialogues,it is important that the researcher
assumptionsand rationalesshould be equally rememberrhat dre timing of the participants'
explicit. contributions has been lost. For example,
Ve want to emphasizehere the advantages between the times when they are writing
as well as the limitadons of any recording or e-mailor other correspondence, the participants
transcribingdecision.It would be possibleto are out in the world doing other things, and
misread what follows as our suggestingthat, each reftrrns anew to the previous message
becauseall data are limited and fallible, no and his or her own reply. So, although the
data are satisfactory, but this is very far from researchercan read the participants' messages
our position. In our view, all discourseis in close sequence,he or she must rememkr
context specific. There is no easy source of (and preferably encodeinto the data) that the
"naturalistic" data in the senseof their being messageswere in fact separatedby time and
unaffectedby the contexts in which they occur many other events.
and the ways in which researcherschoose to As Linell (1982) has pointed out, written
analyze them (Bavelas, 1984, 1995).r The Ianguagediffers from spoken language (and
challengeis to understand these contextual especiallyfrom face-to-facedialogue)in timing
and technicalfactorsand to incorporatethem and many other important respects;a sum-
in a creative manner into one's research. mary of these differences is presented in
Understanding the contextual factors thar Table 4.1. Iflhen individualscornmunicatein
affect the participants' communication can be writing, it is important to rememberthat, even
a substantivegain as well as a merhodological though the printed record of their exchanges
necessity.An understandingof how technical may sometimeslook like a record of face-to-
decisions such as those regarding recording facedialogue,there are crucial differences:For
Discourse Analysis t17

Table4.1 Differences BerweenWritten and Spokcn Language

\\'ritten Text Fdce-to-FdceDialogue

'obiect" Is ephemcral''dynamic'
Is a persisterrt,static
k can be rereadany time' It cannot ordinarily be reviewed.
Ir seldomrequiresa rapid response. The participantsmust respondimmediately,
"on-line.'
Consists of discrete, separatesymbols Is virnrally continuous
Words are easily separatcd. Words and other acts merg€'
Text is organizedspatiallY. Dialogue is organizedt€mPorally.

ls relatively context-free Is highly dependenton context


It usesonly words and punctuation It usesface and hand gesturesas well as
prosodic features.
The words are highly exPlicit. The words can be lessexPlicic.
Text is monologue and solitarY. Dialogue is a "social interplay."
There is no immediatereader. There is an addresseepresent.
The writer and readerare in different places. The participantsare in the same sening.
Text must often be addressedto a gcneral Dialogue can draw on the sertinS,and the
audience. ongoing conversatlon.

ts acquired as secondary socialization Is acquired as Primaty socidization


Literacy is learnedinstitutionally (in schools). Dialogue is learnedinrerpersonally(at home),
Ir is taught with explicit, consciousnorms. It is practicedrather rhan explicitly taught'
The norms are more standardized,with less The norms are freer, with more variation.
variation,
(2000). Copyright 2000 by Sage
SOURCE:Adaptcd from Linell (1982, pp. 5-10); rcprintcd from Bavelasand Chovil
Publications.
especially
NOTE: Some of the features of wrinen text noted here do not apply to computer'mediated communication,
if both partiesare on-line at the samctime.

example, the participants were usually not in any auditory cues' the reader may assumean
rhe samesocial context or physicalsening,and emphasison a different word. Although the
rhey may not have been responding only to wriften version that the resealcher possesses
each other (e.g., memosgo to other recipients may be exactly what the participants wrote
as well; newspapereditorials and lecers to edi- and read, it is important that he or shebe alen
tors have larger audiences).Even the immedi- to unwarranted assumptionsabout hou rhey
are particiPants may have been drawing on wrote and read their messages.Ideally, the
different memories or versions of what they record of their exchangeshould be annotated
previously correspondedabout (or evenof the to remind the analyst constantly of these
messagesthey were currently resPondingto)' factors.
They may have given more or less time to Real-time electronic exchanges(e.g', ICQ'
thinking about and writing or editing their chat rooms) overcome some of the above
replies. The way the recipient (or researcher) limitations, but only if they are videotaped
reads a reply may not be the way the writer (e.g., Garcia & Jacobs, 1999; Phillips 6c
"sid" ig;that is, the writer may haveintended Bavelas, 2000). A prinrout alone cannot
an emphasison one word whereas,lacking capturethe precir timing of rhe exchangesand
118 PERSPECTIVESON INQUIRY

in fact may even misrepresentthe sequencein researcher can come very close to "being
which responseswere acrually made. there." Indeed, one significant advantage of
the telephone format is that in making the
recording, the researcher has potentially
Spoken (Voice-Only) Communication
captured the conversation as it originally
There are many important settings where occurred, ln contrast, for example, speakers
interpersonal cornmunicarion occurs in voice- who leave messageson answering machines
only dialogues. The best example is a tele- are responding in part to their own circum-
phone conversation; conversation analysis stancesand even to the outgoing messagesthey
began with Harvey Sacks's analyses of just heard, which the researcher may over-
recorded telephone conversations, and this look (e.g.,if an outgoing messageis humorous,
format continues to be a fascinatingsourceof unexpected,unclear, or offers limited choices,it
data (e.g., Hopper, 1992). Other examples may affect incoming messagesin systemaric
include answering-machine or voice-mail and interestingways).
exchangesand radio call-in programs. V/e
have usedtelephoneexchangesin the lab when
Face-to-FaceDialogue
we wanted to deal only with audible features
of communicarion rarher than with both audi- Arguably, most interpersonal communica-
ble and visible fearures(Bavelaset al., L990, tion occurs in real-time face-to-facedialogue,
chap. 5). Panicipanrs in spoken records are which, as Table 4.1 shows, has severalunique
more identifiable than those in written records features. Of particular relevance for the
(although lessidentifiable than those in video- researcher'sdccision about recording is the
taped records),so the researchershould obtain fact that face-to-facedialogue has both audi-
permission to use the data. In our experience, ble and visible fearures,so arguably the best
it is a good idea ro usea multipart permission method of recording suchdialogue is by video-
form that allows participants to consentto the taping. However, before we discussthis tech-
researcher'sanalysis of their tapes and still nique, we want to consider other possibilities.
withhold permission, for example, for che Again, the method the researcherchoosesto
resiarcher to play the tapes in public presenta- use for the recording of data should follow
tions where someone might recognize the from and be consistentwith the researchques-
participants'voices. tions he or she is interestedin answering and
Many of the samecautions given above for also with the broader theory that guides the
written communication sdll apply, especiallyif research,
the exchangesexamined are not real-time dia- Most social interaction occurs too quick\'
logues.Even for a telephoneconversation,ir is to be recorded by hand (in writing) as ir
crucial that rhe researcherhear the exchange occurs.It might be possiblefor a researcherto
as itself and not as if it were a face-to-facedia- record the frequency of certain obvious fea-
logue, staying constandy aware of contexnral tures if they do not happen too rapidly or too
facrors rhat affest the panicipants (they are often (e.g., the use of particular words or
not in the samephysical or social context, they phrases),but there is no possibility of accu-
cannot see each other's actions, and so on). rately recording their immediate context or
With this awareness,and if rhe recording of sequence, which most discourse analysts
a real-time telephone conversation is of the would consider essential. Only real-time
same qualiry as the participants heard, the recording will do.
DiscourseAnalysis 179

Audio recording of face-n-face dialogue.l\e cautions, however, about accepting this


use of audio-only recordings in discourse assumption immediately.
analysis is common and offers severaladvan- First, in face-to-facedialogue, the partici-
tages. The equipment can be relatively inex- pants often use deictic expressions,such as
pensive and small, making it portable and *over here," "that oner" or'like this." In such
unobtrusive; convenient standard trans- casesthe words are insufficient, becausethey
cription equipment is available that allows were intended by the speaker to bc supple-
repearedplayback; and, as noted eadier, audio mented by a visible action such as looking,
recordings provide a degreeof anonymiry for pointing, or gesturing.
panicipants. However, one nuy need rela- Second,the meaningof a polysemousword
tively sophisticatedequipment to ensure rhat can be specified by a nonverbal demonstra-
voice qualities are accuratelyreproduced,that tion. In Kendon's(1985) example:
all voicesare equally audible, and that over-
lapping speechcan be disentangled. I havea Minolta SLR [still-photo]camera,
The central issue (and pot€ntial contro- and wheneverI referto this in my family I
versy) concerning rhe use of audio recordings always refer to it as "my Minolta."
Recently,I acquired a Minolta super-8
of face-to-face dialogue is whether such
movie camera.Consequently the word
recordings are necessarilyincomplete-and "Minolta" becameambiguous, for it could
not iust for researcherswhose primary inter- referto either[cameraj.In the courseof a
est is in nonverbal communicarion. Unlike convcrsation with my son--soonafter this
telephone conversants, participants who new camerahad beenacquired-hesaidto
interact in person are in the same physical me: "You could do it with yours, your
and social sefting and able to seeeach other Minolta." As hesaid"Minolta" he liftedhis
handsup, thumb and forefingerof each
as well as the same obiectsor eventsaround
hand extendedat right angles to one
them. Each can seehow the other looks, both anotherand heldon cithersideof his face,
can refer to features of the environment therebymodelingthe action of holding a
that they can see,eachknows when the other [still-photoJcamera.. . . By doing this he
person is looking ar him or her, both can disambiguated *Minolta," clearlyindicating
seeand interpret any nonverbal actions,and that it wasthe SLRhe was referringto and
so forth. not the moviecamera.(pp.225-2261
There may be instances where the
researcher's interestis entirely lexical, that is, In this case, the hand gestures specified the
focusedon verbal or grammatical features.Or meaning of the word. In other cases, the
the researcher'stheory may hold that any speakermay complete a verbal phraseentirely
visible features(e.g.,gaze, body orientation, with a gesture or facial display rather than
hand and facial gestures, or the physical finding the word ("mixed syntax"; Slama-
obiectsor eventsin the setring)are noncom- Cazacu,1976), so that an observerneedsboth
parts to understand the ufferance. Kendon
municative, redundant, or irrelevant. Thus the
(1985) describesa host who said to his guest,
decision to audiotape a face-to-facedialogue
is part of a cleartheoreticalposition and must whom he neededto drive home,
be recognizedas such. If the phenomenon or
"Last night I had morecoffeethan usual. . .
processunder study can occur only audibly
and I didn't sleepwell, so maybewe oughta
and could not be affected by what the /GESTLTRI/."In this gesturehe placedhis
researchercannot see,then the use of audio two extendedindex fingersside by sideand
recording is a good option. !7e offer several thenextendedboth armsawayfrom himself
120 PERSPECTIVESON INQT.nRY

and upwardsin the directionof the door, theoretical choices could be resolved empiri-
He thereby clearly indicatedthat he and cally. Much more researchis neededthat com-
[hisguest]shouldleave.(p.224) pares audio-based and video-based analyses
of the same data. Those who advocate video
Third, some aspectsof communication are recording should suppoft their position by
typically not put into words. In multiparty showing empirically that it matters. For
interactions, whom rhe speaker is addressing example, we found in our own work thac
(and who is anending to the speaker)is rypi- speakersoften made certain g*t:rrtes (inter
cally indicated by gaze, rarher than verbally actiue gestures)that elicited responsessuch as
(Vertegaal,1998). Finally, an audio recording back channels from the listener, yet SOyoof
will not capture what happens in the ubiqui- these gestureswere completely nonredundant
tous (auditory) pauses.A 'pause" on audio- with speech(Bavelas,Chovil, Coates, & Roe,
tape is only a verbal pause.It is not likely that 1995; Bavelas Chovil, Lawrie, & Wade,
the participants were momentarily frozen in l992l.That is, there was no verbal indication
time, doing nothing; they may have been of the speaker'saction and meaning, so the
smiling at each other, gesturing,avoiding each speaker'sinfluence on the listener's response
other's gaze, or engaging in any number of would be ignored, inexplicable, or spuriously
other communicative acts. Researcherswho interpreted from an audio-only recording.
interpret (auditory) pausesas, for example, A later study demonstrated the interaction
hesitadon or reluctance might have reached between speaker gaze and listener respons€s
enrirely different interpretations if they had (Bavelas, Coates,& Johnson,2002;this anicle
worked from video recordings. (fu will be is accompaniedby a video on CD that lets the
seen below, we are by no means suggesting readerseethe effect).
that a video recording missesnothing.)
The reader may appreciareby now why we Video recording of face+o-face dialogce. ln
have proposed thar this issueis both theoret! most of this subsection we discuss issues
cal and porentiallycontroversial.Researchers relatedto researchers'making their own video
who insist on videotaping face-to-face dia- recordings. We want to note, however, that
logue (as our researchgroup does) often hold researcherswho find that this is not possible,
the theoretical position rhat visible feailres or who want to examine wider rangesof con-
such as hand and facial gesturesor gazeare an texts than are directly available to them, may
integral pan of communication (Bavelas& want to consider the possibility of analyzing
Chovil, 2000). Researchers who preferaudio- documentaryfilms. For example,the National
tapes often hold the theoretical position that Film Board of Canada specializesin high-
thesevisiblefeaturesare not so important as to quality documentaryfilms, hundreds of which
iustify the expenseand inconvenienceof video are available through the board's Web site at
recording. We obviously cannot seek to www.nfb.ca. We have also used training or
resolve this issue here, but we suggest rwo demonstrationfilms of psychotherapybecause
guiding principles. First, the theoreticalchoice of the ethical problems of filming actual ses-
should be explicir and considered. Research- sions (Bavelas,McGee, Phillips, 6r Routledge,
ers who chooseaudio-only recordingshould 2000; Phillips,1999).
explain in their researchreporrs their underly- It is important to say at the outset that
ing theoretical position and rationale as it researcherswho study videotapesshould have
applies to the particular studies at hand: the no illusions that viewing such recordingsis the
same burden falls on proponents of video. same as being there, As we note below, even
Second,many of the issuesthat guide these when a video recording is made with the best
Discourse Analysis 721,

equipment, it is a smaller-thanJife, selective, more options for later storageand analysis.In


rwo-dimensional image that cannot in any any case,a permanenrtime signal on the tape
casecapture everything the participants could is invaluable for later locating and identifying
perceive(e.g., temperature, smells, and sights segments.
and sounds our of range of the recording I7ith whatever equ.ipment,camera place-
equipment). ment is an important decision:lVhat you seeis
Video recordings also present researchers what you get. A video recording will include
with the most challenging ethical issues, only the view and focus the rcsearcherselects.
becauseparticipants are clearly identifiable to Studies of social interaction usually require
anyone who knows them. If the participants' that the behaviors of both participants be
facial features and voice qualities are not of captured by bofi lens and microphone. (Choos-
interest, they can be electronically scrambled, rng to film only one of the panicipants would be
but this is usually not the case.[n our own a major theoreticaldecision.)The field researcher
research,we use a multipart panicipant per- is usually limited to only one cannera,which must
mission form that lets each panicipant (after capftre everyone of interest; this probably
having viewed his or her tape) give or with- meansa fixed perspecriveand no close-ups.If
hold permission separately for keeping the the participants are moving around, the
rapeat all, analysisby the researchtearq view- camera must track them or be set for a wide-
ing by audiencesat other universities or at angleshot. When there is a choice of focus, the
conferences,viewing in classesat our univer- researchermust usually choose, for example,
srry, and any other uses we can reasonably between being far enough away to see hand
.rnricipate(e,g., on our Web site). Then we gesturesand being close enough to see facial
transfer the coded permission starus to all expressions.
copiesof the tapes. ln the lab, it is possible to use more than
One potential ethical decision, whether or one camera; the shots from two or more
not to hide the camera in the first place, is camerascan be configured and fed onto one
becominga nonissue.lUith more experiencein split-screen videotape by a special-effects
vrdeotaping interaction, researchershave dis- generator or "splitter." It is also possible to
-olered that filrning openly does not produce s€atthe participants in preselectedplaces.We
-arrificial' will illustrate some of the possibilities wirh a
data. For example, Goodwin
l98l ) filns openlyin the field, aswe do in the two-person interaction.'With rwo cameras,the
ieb. Certainly, presentingthe cameraas a non- researchercan capture a front view of both
;hreareningaspectof the study is essential,but pardcipants as they looked to each other, by
se also have popular rcchnologyon our side: seatingthem almost face-to-face(but not per-
\losr parricipants are now used to being fectly, so they do not block each other's cam-
r rdeotapedin banks, stores,and airpons, and eras) and placing the qamerasopposite each
::r.rnvhavetheir own camcorders. other, behind eachparticipant. This configura-
Researchers who chooseto record on video tion aims a camera over the shoulder of each
::rusrconsider many factors prior to collecring person, dirrctly at the face of the other, so the
:ic' data (Goodwin, 1993). First, they must viewer seesboth individuals as they looked to
-iroose a format to record in. Popular choices each other. Then, choosing a vertical split pro-
irc VHS, SuperVHS, 8mm, and digital. duces a recorded image that looks as though
.\n.rlog camerasand camcordersare relatively the participants were side by side, which is
-ieap. Digrtal video is more expensive,but ir somethingthe analyst must adiust to. (It is also
':iers improved image quality, does not possible to produce a rwo-way split-screen
.:sr:r.rdewith repeatedviewing and provides effect with iust one cameraby placing a mirror
122 PERSPECTTVES
ON TNQUIRY

behind one of the participants, provided this provide similar and familiar ease of access,
does not distract the other participant.) If they are often more convenient to use than
there is a third cameraavailableand a suirable audio or video recordings. That is, perusing
splifter, the third camera can capture both audio or video recordings and locating par-
participants from the side with a wide-angle ticular segmentscan be awkward and time-
lens. This third image can then be split onto consurning, so many researchers prefer ro
the bottom of the recorded image, where it work with transcriptsinstead.Note, however,
creates a pseudo-three-dimensionalperspec- that the advent of digitization may be reducing
tive, in that the analyst can see the panici- the transcript's advantage of convenience
pants' spatial movements, including hand becausethe linear viewing constraint does not
gesfuresand their movementstoward or away hold. With a digitized video (even of an
from each other. It can take an analyst a while originally analog recording) and appropriate
to get used to this view and be able to recon- sofrware,' it is possible to go immediately ro
struct from it mentally the three-dimensional specified s€gmentsand to view them repeat-
interaction; it is certainly not a literal re- edly with a simple command. Segrnentsthat
creation of the interaction. seem to demonstrate the same process or
Finally, whether recording by audio or phenomenon can be stacked up together for
video, researchersmust rememhr the pracd- comparison. (Digitized video is also easier to
cal wisdom that all of us have learnedthe hard transcribe,becausethe video image and word-
way. First, set up the equipment and try it out, processingprogram can be tiled side by side on
from a trial recording through analysisof that the computer monitor, rather than using
recording. It is much better to find out that severalpiecesof equipment.)
you cannot analyze one trial recording The other reasonsfor transcribing audio or
(becausethe sound is inaudible, the image is video records are increasingly theoretical, in
too small, you did not include somerhingthat that they involve assumptions(either implicit
had not seemedimportant at the time, and so or explicit) about the nature of the discourse.
on) than to find our after having recorded all The researcher may sirnply produce a verbal
of your data. And second, technophiles and transcript becausehe or she rnrplicrdy assumes
rechnophobesalike can solve the vast majoriry thar words are the significant ftature of commu-
of equipment problems by asking three ques- nication. Or the researcher may be explicidy
tions: Is it rurned on? Is ir plugged in? Is it organizing the &a according to his or her theo-
connected? retical interests, and a specific kind of nan-
scription is the first step. In either case, the
effecs of nanscription can h made clear by an
Transcribing
analogy with furmulations in conversation (in
Why rranscribe?There are several reasons which one person summarizesor describeswhat
that researchersneed to consider separately the other person has just said; Garfinkel &
before embarking on this time-consuming Sacks, 1970). Heritage and Watson (1979)
activity. The first is easeof access.Consider point out that formulations inevitably preserue
that, with textual materials (such as e-mails or something of the original st^rcment, delete
leners), the analyst can work directly with the pan of it, and transform other parts. A tran-
printed data and can thereforeeasilystudy any scripdon is like a formulation; it preserves
part of the data in any order and in any time some aspects(e.g., the words), deletesothers
sequence.Becausewritten transcripts of non- (e.9., visible features), and rransforms others
textual data (e.g., telephone conversations) (e.g.,prosodic featuresmay be representedin
Discourse Analysis 123

words or symbols). The challenge to the review, see Dressler & Kreuz, 2000). The
researcheris to be aware ofthesc changesand transcripts these researcherscreate therefore
to be explicit about his or her rarionale. include precise description and timing of
Many established transcriprion sysrems thesephenomena(e.g,,"u:h" meansan elon-
exist (e.g.,Jefferson,1985). The method the gated sound). Researchersworking from a
researcherchoosesto us€ in transcribing the functional perspective,on the other hand, will
data-whether a merhod developedby some- emphasize the rhetorical or meaningful
one else or by the res€archerhim- or herself- impacts that behaviorshave in a conversarion.
determineswhat will be preserved,what will These researcherswill be les interested, for
be ignored, and how the data are transformed. example, in the absolure duration of a pause
Choosing a transcription method or crearing than in an interpretive evaluation of what that
one is therefore an important theoretical deci- pause means in a conversation (Gumperz &
sion and should be treated as such. Several Berenz,1993), so a pausemight be transcribed
papers have been wrirten on transcription with an annotation marking it as an indication
(e.g., Edwards, 1993; Ochs, 1979), so we of uncertainry.
cover only the main considerationshere,organ- With a video record, the researcherhas a
ued into content (what to include) and format. great deal more to decideabout, including how
As you consider the following, keep in mind to transcrib€ hand gestures,facial displays,
that not transcribing is also an option. gaze direction, and postural and limb posi-
tions. If tfiese are included in the transcripr,
how should they be represenred?One impor-
What to Include?
tant choice is beween a purely physical
First, the researchermust consider his or descripdon and a meaning-baseddescription.
her own researchgoals. Transcription is an For example,in Kendon's "Minolta" example
extremely time-consuming process-an hour above, he first describesthe physical actions
or more for each minute of dialogue would and then adds information about what those
not Lleunusual-so it is wasteful to transcribe actionswere meant to depict ('holding a [still-
more (or less) information than one needs. photol camera"). Ekman and Friesen's(19781
Most researcherstranscribe all of the words Facial Action Coding System(FACS)describes
spoken, but practices vary. lf a researcheris facial expression by referring to rhe muscle
interested only in where listeners' responses groups involved-a purely physical descrip-
occur (Bavelas,Coates,& Johnson,2000), for tion, In contrast, Chovil (1989, 1997-19921
example, he or she might transcribe all of the has developed a meaning-based sysem in
listeners' responsesfrom the videotape but which the analystdescribeswhat the facial dis-
transcribe the speakeronly around the points play is depicting (sadness,skepticism, and so
rvhere listener responsesoccur, Such a Ean- on). It is wonh pointing out that Chovil's data
script would be unusable as a version of the demon$ratehigher interanalystreliability than
speaker's contribudon, but that is not its the FACS. The same can be true for interpre-
purpose, tive approachesto gestures(for funher descrip-
Whether one is working with an audio or tion of meaning-basedsystemsfor describing
video recording, the prosodic and other nonverbal acts, see Bavelas 6r Chovil, 2000,
paralinguistic fearures are available to be pp. 175-183).This agreementis encouraging
included, Many researchers,for example, are becausesuchapproachesare more suitedto the
interested in phenomena such as intonation, goal of discourr analysis,which is to study
srress,pauses,and timing of exchanges(for a communicationat the level of meaning,
124 ON INQUIRY
PERSPECTTVES

Format Although there are many similarities among


different transcription systems,there are also
The format of the transcript will influence
many differences, reflecting the differing
the analysis carried out by the reader.
goals,theorericalassumptions,and sometimes
Transcripts should be formatted so that they
unrecognized presuppositions of different
are easyto read and the appropriate informa-
researchers.For example, an important for-
tion can be extracted quickly. To make tran-
matting choice concerns how to arrange
scripts readable, the researcher should
speakersinto turns, Notice that this format
consider both the visual prominence and the
presupposesthat alternating speaking turns
spatial arrangement of the data (Edwards,
are a basic unit of dialogue, which rnakes
1993). Visual prominencerefers to how the
simultaneousor overlapping speechproblem-
data appear, such as the use of italic or bold-
atic and igrroresthe simultaneous, noninter-
face type, underlining, and font size.Using any
ruptive facial expressionsand nods of those
one or a combination of thesetools can give
who are not speaking.There are at lea$ three
prominence to particular elements of the
different arrangementsused in existing tran-
transcript. [n addition, data elementscan be
scription systems:verticd, column, and pani-
positioned spaciallyon a transcript so as ro
ture (Edwards, 1993), as shown in the
improve both readabiliry and efficiency.
following examples(which come from Tannen,
Note that the spatial format of the tran-
1989, p. 60). The uertical format is perhaps
script also reflects often unrecogrrizedtheo-
most common. It is similar to the script of a
retical assumptions.For example, becauseof
play, but usually with overlapping speech
Western writing conventions, behaviors for
shown in aligned squarebrackets:
which descriptionsare placed either to the left
or above other descriptionsof behaviors in a
A: Like he saysthat he saysthat Arnerican[s]
transcript are seen as occurring before (and
potentially as causesof) behaviorstranscribed B: [Yeah]
to the right or below them (Ochs,1979). A: or Westcrnerstendto be u:h think of the
Severalsimilar featurescan be found across bodyand the soul as two differentthfings,]
different transcription systems.Fvst, related B: lright]
euents tend to be located near each other. For
example, descriptions of gesturesare usually As Ochs (1,979)points out, this format has
locatednear the speecheventswith which they several effectsr It encourages the reader to
co-occur.Similarly, transcriptionssystemsthat link adiacent utterancesand to see each as
include prosodic information (e.g., the successively contingenton those precedingit;
London-Lund Corpus; Svartvik & Quirk, it also tends to make the reader see partici-
1980), such as prominent syllable stress, pants as equally involved and influential in
will put that information with the relevant the interaction.
syllables.Second,visualprominenceis usedto When the researcherwants to highlight the
separate qualitatively different data. Trans- influenceof one personon rhe other, a colwnn
cribers' comments are usua[y bracketed or format may be more appropriate:
presented in font that is different from that
used for the data. Third, as noted above, SpeakerA SpeakerB
temporal rektionships are usually presented Like he saysthat he says
in a left-to-right and top-down format, so thar thatAmericans... Yeah.
earlier eventsare presentedbefore later events. 'Westerners
or tend to be u:h . . .
Similarly,explanatorymaterialis placedat the think of the body and the soul
beginning,where it frames all that follows. as two different things. Right.
Discourse Analysis 125

In this format, the speakeron the left is given possible.For example, one can underline the
prominenceover the speakeron the right; thus words that overlap the gesture and describe
SpeakerA may be implicitly s€enas initiating the gesture in brackets underneath. The fol-
the behaviors of SpeakerB. This format does lowing example(adaptedfrom Bavelas,1994,
not provide an easy way to indicate precisely p. 2071also illusrates meaning-baseddescrip-
when the speakersoverlap, as they often do in tion of the gestures;the speakersare discuss-
spontaneous dialogue. However, it might be ing the Royal Canadian Mounted Police
suitable for capruring asymmetriesin dialogue (Mounties,or RCMP):
or for depicting parallel speech (e.g., when
both are trying to get the floor or are having a A: Is-the only RCMPin the area
shoutingmatch). is-whatl Colwoodor?
Partiture formats (Tannen, 1984) are [drawsjaggedcircles,
desigrredto caprure interaction in which there
asif on a mapl
may be rnany instancesof simultaneousbehav-
iors. They are similar to multipart musical B: Well rhat's,that'sa, a
rggulardetachmentin Colwood,
scoresand preserveboth timing (horizontally
as opposed to venically) and the assumption Idrawstightercircle]
of conversationalequaliry:
A final, pracrical warning: The transcrip-
A: . .. American[s] tion process rakes a toll on audio or video
B: records. Repeatedly playing, pausing, and
[yeah]
rewinding an analog tape will soon degradeits
A: or Westerners
...
qualiry. Therefore, researchersshould always
work from copies, using the originals only to
This format is continued, with line breaks as
make new copies.As noted above,one advan-
needed.
rage of digital records is that they will nor
Another imporrant choice the researcher
degrade; however, it is still wise to make at
must make when creating a transcript is where
least one copy.
to put visible elementsof the dialogue,such as
hand gestures and facial displays, if indeed
New Technologies
they are syttematicallyincluded.Somesystems
separatethe visible elementsby placing them Although most of us develop or adapt our
in their own column on rhe right-hand side of own systemsfor managing our data and cod-
the transcript, roughly parallel to the spoken ing social interactions, commercial sofrware
words in the left column. This formar has sev- systems designed specifically for these msks
eral effects:First, it imposesa theory of verbal are available. One of these,Noldus Observer
and nonverbal (audible and visible) acts as Pro, is a systemfor collecting,analyzing,and
separatechannelsand giveswords prioriry by presentingobservationaldata (seethe ITeb site
placing them on the left. Second,the relation- at www.noldus.com).It can be usedro record
ship betweenthe rwo is lost, For example,the behaviorssuch as postures,movements,posi-
use of a gesture to supplement words is less tions, and facial expressions.The analyst uses
clear, and the precise verbal context of the key pressesto log events and the times at
gesturemay be lost. which they occur. This systemmakes ir possi-
Researcherswho rake an integratedview of ble for the researcherro work directly from the
audible and visible acts would use a format video record rather than from a written tran-
that integrates the verbal and the visible and script of it. It also allows the analyst to add
also preserves their timing as precisely as notes and conrmentsthat are then stored with
126 PERSPECTIVESON INQUIRY

the video data. The system providc facilities from theoreticaland philosophical debatesto
for descriptive staustics,lag sequentialanaly- technical considerations,As on any tour, we
sis, and reliability analysis.The data can be could have stayed longer in each place, and
exponed into spreadsheets,databases,or sta- interested readers may want to know rnore
tistical packages.The Noldus Observerpack- than we have provided here or in our refer-
age (which costs severalthousand dollars for ences.Van Diik's (1997a,7997b)two-volume
the sofrware alone) can be purchasedeither introduction covers the many kinds of
as software only or as part of a complete discourseanalysis in much more detail, and
hardware/softwarebundle that includesanalog each chapter has a list of further readings on
or digital cameras,VCRs, video caprurecards, its topic. Finally, many journals (such as
and time-code generators. Transana, on the Discourse Processes,Dkcourse and Society,
other hand, is a free tool for transcription and Researchon Language and Social Intetuction,
qualitarive analysisof videotape (seethe Web and the Journal of Language and Social
site at www2, wcer.wisc.edr:./Transana/index- Psycbologyl include srudies that have used
html). some form of discourse analysis. Interested
Severalaffempts have been made, such as readerswill find that locating such anicles in
the Texr Encoding Initiative (TEI), to intro- their own areasof interest is an excellentwar'
duce guidelinesfor electronictext encoding to continue the tour and to decidewhere (and
and interchange (see the TEI Web site at if) they want to go on their own.
www.tei<.org). The goal of thesesystemsis to
create a common encoding schemefor tran-
NOTES
scripts and other text, rhat is, to reduce rhe
diversiry of existing encoding practicesand to 1. Many readerswill be stardedto seedata from
encouragethe sharing of data, as well as to a psychology lab included as "naturally occur-
make it possible for the data to be read by ring." However, we agree with khegloff (1992.
machines.The TEI is one of the only system- p. 116) that, as long as the participants are nor
confederates and are interacting spontaneouslv.
atized attempts made thus far to develop a
even within an assigned task or topic, their dis-
fully general text encoding model and encod- cours€cannot be arbitrarily dismissedbecauseoi
ing conventions. Becauserheseguidelines are the context in which it occqrred.
meaht to be applicable over a wide variery of 2, It is undoubtedly because of the inductitr
data, they are necessarilyrestrictivein terms of preferencesof many discourse analysts that the.
what they permit the user to do, as well as eschewthe term coding, which has implications or
being time-consuming to learn. We would an a priori systemthat the term analysisdoes nor.
expect the sametrade-off to apply ro any new 3, Ironically, the term natualistic actuall.
'as if
system-+hat is, the user must choor between means narural" {e.g.,a naturalisticpainting .
the porential hnefits of a common systemand 4, We are currently using the softwar.
its methodological constraints, which, as we Broadway (www.b-way.com) becausc of irr
have emphasized throughout, also impose higher-quality AM format and becauseit allos'
implicit theoretical assumprions. us to create"loops," that is, to isolateand replar
selectedsegmentseasily.

WHERE TO FIND MORE REFERENCES


To retum to our inrroductory metaphor, we
Atkinson,J. M., 6( Heritage,J. (Eds.).(198.t
have traveled through a lot of territory in the
Structurcsof socialaction:Studiesin conve,.
far-flung field of discourseanalysis and have sation analysis. Cambridge: Cambridg.
stoppedto look briefly at many of its fearures, UniversityPress.
DiscourseAnalysis L27

Bartlett, F. C. (1932). Remembering.Cambridge: Brown, G. (1995). Speakers,Iistenersand commu-


CambridgeUniversityPress. nication: Explorations in discourse analysis.
Bavelas,J. B. (1984). On "naturalistic" family Cambridge: Cambridge Universiry Press.
research. Family Process,23, 337-341. Caldas-Coulthard, C. R., & Coulthard, M.
Bavelas, J. B. (1987). Permitting creativity in (1996a1. Preface. In C. R. Caldas-
science, In D. N. Jackson 6( J. P. Rushton Coulthard & M. Coulrhard (Eds.),Texts and
(FAs.l,Scientificexcellence:Origins and assess- plactices: Readingsin critical discourseaflaly-
ment (pp.307-3271.Newbury Park, CA: Sage. sl's(pp. xi-xii). London: Routledge.
Bavelas,J. B. (1991). Someproblemswith linking Caldas-Coulthard, C. R., & Coulthard, M. (Eds.).
goals to discourse. In K. Tracy (Ed,l, Under- (1996b). Texts and practices: Readings in
standing face-to-face interaction: Issues critical discourse analysis. london: Roudedge,
Iinking goals and discourse (pp. 119-130). Chovil, N. (1989). Communicatiuefunctions of
Hillsdale, NJ: lawrence Erlbaum. facial disphys in conversatioz. Unpublished
Bavelas,J. R. (19941. Gesruresas pan of speech: doctoral dissertation, University of Victoria,
Methodological implications. Research on Department of Psychology.
Language and Social hteraction, 27, 201-221. Chovil, N. (1991 -19921.Discourse-orientedfacial
Bavelas,J. B. (1995). Quanritativeversusqualita- displays in conversation. Researcbon Langu-
tive? In W. Leeds-Hurwitz (Ed.), Social age and Social Interaction, 25,763-194,
approaches to commrnication (pp, 49-62lr, Clark, H. H., & Wilkes-Gbbs,D. (1985),Referring
New York: Guilford. as a collaborative process. Cognition, 22,
Bavelas,J. 8., Black, A., Chovil, N., & Mullett, J. 7-39.
(1990). Equiuocal communicatioa.Newbury Coates, L. J. (1991). A collaboratiue theory of
Park, CA: Sage. inuersion: Irony in dialogue. Unpublished
Bavelas, J. B., 6. Chovil, N. (1997). Faces in master's thesis, University of Victoria,
dialogue. In J. A. Russell E<J.-M. Fernandez- Depanment of Psychology.
Dofs (Eds.), The psychology of facial expres- Coates,L. J. (1997). Causalanributions in sexual
sioz (pp. 334-3461. Cambridge: Cambridge assault trial iudgments. Jountal of Language
University Press. and SocialPsychology, 16, 278-296.
Bavelas,J. 8., & Chovil, N. (2000). Visible acts of Coates,L. J. (1998). Methods for deuelopingscoring
meaning: An integrated messagemodel of lan- slsteffis and calculating relability in discourse
guage use in face-to-facedialogue.lourul of analysis, Unpublished manuscript, Universiry
Ianguage and SocialPsychology, 19,163-194. of Victoria, Department of Psychology.
Bavelas,J, B., Chovil, N., Coates,L.J,, 6( Roe, L. Dressler, R. A., 6( Kreuz, R. J. (2000).
(1995). Gestures specialized for dialogue. Transcribing oral discourse: A survey and
Personality and Social Psychology Bulbtin, model system.DiscourseProcesses,29,25-36.
21, 394-405. Edwards, D., & Middleton, D. (1987). Conver-
Bavelas,J. 8,, Chovil, N., lawrie, D. A., & Wade, A. sation and rememberingr Banlem revisited.
(1992J, Interactive gestures. Discourse App lied Cognitiue Psychology, 1, 77-92.
Processes,15, 469-489. Edwards,J. A. (1993). Principlesand contrast-
Bavelas,J. B., & Coates,L. I. {19921.How do we ing systems of discourse transcription. In
account for the mindfulness of face-to-face J. A. Edwards & M. D. Lampert (Eds.),
dialogue? Communication Monographs, 59, Talking data: Transcription and coding in
301-305. discourseresearch1pp. 3-32). Hillsdalc, NJ:
Bavelas,J. B., Coates,L.J., & Johnson,T. (2000). Lawrence Erlbaum,
Listeners as co-narrators, Journal of Ekman, P. (1993). Facial expressionand cmotion.
Personality and Social Psychology, 79, American Psychologist, 48, 384-392.
941-952. Ekman, P., & Friescn,W. V. (1978). The Facial
Bavelas, J. B., Coates, L. J., & Johnson, T. Action Coding System: A technique for
(2002'1,Listener responsesas a collaborative the n easurement of facial mouement,
activiry: The role of gaze.lournal of Com- Palo Alto, CA: Consulting Psychologists
munication, 52, Press.
Bavelas, J. 8., McGee, D., Phillips, 8., & Fairclough, N. (1992). Discourse atd social
Routledge, R. (2000). Microanalysis of change. Cambridge: Polity.
communication in psychorherapy. Human Fairclough,N. (1993). Critical discourseanalysis
Systems,11, 47-66. and the markctization of public discourser
128 PERSPECTIVESON INQUIRY

The universiries.Discourse and Sociery, 4, Kendon, A. (1985). Some uses of gesture.ln


13 3 - 16 8 . D. Tannen 6c M. Saville-Troike (Eds.),
Fowler,R. (1996).On critical linguistics.In C. R. Perspectiues on silence (pp. 215-2341.
Caldas-Coulthard6< M. Coulthard (Eds.), Norwood, NJ: Ablex.
Texts and practices:Readingsin critical dis- Kraut, R. E., & Johnston,R. E. (1979).Socialand
caurse analysis (pp. 3-1,41. London: emotionalmessages of smiling: An ethologi-
Routledge. cal approach. Journal of Personality and
Garcia,A. C., 6( Jacobs,I. B. (19991.The eyesof SocialPsychology,37, 1539-l553.
the beholder:Understandingthe turn-taking Leeds-Hurwitz,W. (79871.The social history of
system in quasi-synchronouscomputer- the Natural History of an Interview:A multi-
mediated communication. Research on disciplinaryinvestigationof socialcommuni-
Languageand SocialInteraction, 32, 337-367. cation. Research on Language and Social
Garfinkel, H., Ec Sacks,H. (1970). On formal Interaction,20, 1-51.
structures of practical actions. ln Linell, P. (1982).The urinen languagebiasin lin-
J. C. McKinney & E. A. Tiryakian (Eds.), guistics.Linkoping, Sweden:University of
Tbeorctical sociology (pp. 337-3661. Linkoping,Departmentof Communication.
New York: Appleton-Century-Crofts. Linell, P. (19981.Approaching dialogue: Talh,
Gifbert,G. N., Ed Mulkay, M. (1984).Opening interactionand contextsfu dialogicalpercpec-
Pandora's box: A sociological analysis of tizes.Amsterdam:John Benjamins.
scientists'discourse.Cambridge:Cambridge Nunan, D. (1993).Introducingdiscourseanalysis.
UniversiryPress. London: Penguin.
Goodwin, C. (1981). Conuersationalorganiza- C)chs, E. (19791. Transcription as theory. In
tion: lnteraction betweenspeakersand hear- E. Ochs & B, B. SchieffelinlEds.l,Deuelop-
ers.New York: AcademicPress. ttrental pragmatics (pp. 43-72). New York:
Goodwin,C. (1993).Recordinghumaninteraction AcademicPress.
in naturalsettings.Pragmatics,3,181-209. Parker, l. (1992ll. Discoursedynamics: Critical
Gumperz,J.J., & Berenz, N. (1993).Transcribing analysisfor social and indiuidual psychology.
conversational exchanges. InJ. A, Edwards& London: Routledge.
M. D. Lampert (Eds.), Talking data: Phillips,B. (19991,Reformulatingdisputenarra-
Transoiption and coding in discourse tives through active listcning. Mediation
research(pp. 9l-1221.Hillsdale,NJ: Lawrence Quarterly,17, 157-180.
Erlbaum. Phillips, B., & Bavelas,J. B, (2000, June).
Har16, R. (1995). Discursivepsychology.In Comparing collaboratiueuersusautonomous
J. A. Smith,R. Harr6, & L. van Langenhove modelsof conuersationin computer-mediated
(Eds.),Rethinkingpsychology(pp. 143-159). communicatioa.Paperpresentedat the annual
. Thousand Oaks,CA: Sage. meeting of the International Communi-
Heritage,J., & Watson,R. (19791.Formulations cationAssociation,Acapulco.
as conversational objects.In G. Psathas(Ed.), Potter, J. 0996lr. Representingrealiry: Discourse,
Eueryday language:Studiesin ethnometbod- rhetoricand socialconstructioa.l,ondon: Sage.
ology (pp. 123-1621.New York: Irvington. Potter,J., & r0iletherell,
M. (1987),Discourseand
Hodge, R., & Kress, G. (1993). Languageas social psychology: Beyond attitudes and
ideology (Rev. ed.). London: Routledge& behauiour.London:Sage.
KeganPaul. Schegloff,E. A. (1,9921.On talk and its institu-
Hopper, R. E. (1992). Telephoneconuersation. tional occasions.In P. Drew & J. Heritage
Bloomington:IndianaUniversiryPress. (Eds.l, Talk at work: Interaction in institu-
Jefferson,G. (19731.A caseof precisiontiming in tional settings(pp. 101-134).Cambridge:
ordinary conversation: Overlapped tag- CambridgeUniversiryPress.
positionedaddresstermsin closingsequences. Schiffrin, D. (1994\. Approachesto discourse.
Semiotica,9, 47-96. Cambridge,MA: Blackwell.
Jefferson,G. (1985).An exercisein the transcrip- Schober,M., & Clark, H. H. (1989). Under-
tion and analysisof laughter.In T. A. van standing by addresseesand overhearers.
Dijk (Ed.) Handbook of discourseanalysis: CognitiuePsychology,21, 27L-232.
Vol. 3. Discourseand dialogue(pp. 254a1. Scon,M. B., 6{ Lyman,S. (1990).Accounts.In
London: AcademicPress. D. Brissett& C. Edgley(Eds.),Life as tbeater
Discourse Analysis
129

A dramaturgical sourcebook (2nd ed., van Dijk, T. A. (1997c).The study of discourse.In


pp. 219-2a2). New York: Aldine de Gruvter. T. A. van Diik (Ed,), Discourse studies:A
Slama-Cazacu , T. (1976\, Nonverbal components multidisciplinary introduction: vol. L
in messagesequencer"Mixed syntax," In Discourse as st/ucture and process(pp. 1-34).
W. C. McCormack & S. A. !(/urm (Eds.), Thousand Oaks, CA: Sage.
Language and man: Anthropologkal issues Vertegaal, R. (1998). Look utho's talking to
(pp. 217-2271.Thc Hague: Mouton. u.,hom. Enschede,Netherlands: Universiw of
Stubbs, M. (1983). Discourse analysis: The socio- Twente, Cognitive ErgonomicsDepanment,
linguistic analysis of natural language. Watzlawick, P., Beavin Bave.las,J., 6c Jackson,
Chicago: University of Chicago press, D.D. (1967't.Pragmaticsof human commu-
Svarwik, J., & Quirk, R. (1980). A corpus of nication: A study of interactional patterns,
Englisb conuersation.Lund, Sweden:Gleerup. pathologies, and paradoxes. New york:
Tannen, D. ( | 984). Conversational stl[e. W. W. Norton.
Norwood, NJ: Ablex. Willig, C. (1999). tntroduction: Making a diff.er
Tannen, D. (1989). Talking uoices: Repetition, ence. In C. Willig (Ed.l, Applied discourse
dialogue and imagery in conuersational dis- analysis: Social and psychological interuen-
course. Carnbridge: Cambridge Universiry tions (pp. l-21). Buckingham:Open Universiry
Press. Press.
Thomas, L. 11974ir.The liuesof the cell. New york: Wood, L. A., & Kroger, R. O. (2000). Doing dis-
Bantam. course analysis:Metbods for studying action
van Dijk, T. A. (Ed.). (L997a1.Discoursestudies: in talh and !ext. Thousand Oaks, CA: Sage.
A multidisciplinary introduction: Vol. I. Yngve,V. H. (1970). On gening a word in edge-
Discourseas structule and process.Thousand wise. ln Papers from the Sixth Regional
Oaks, CA: Sage. Meeting of the Chicago Linguistics Society
van Dijk, T. A. (Ed.). (1997b), Discourse studies:
{pp. 567-5781.Chicago: Chicago Linguistics
A multidisciplinary introduction: Vol. 2. Sociew.
Discourse as social interaction, Thousand
Oaks,CA: Sage.

S-ar putea să vă placă și